rspb.royalsocietypublishing.org
Research Cite this article: Bucciarelli GM, Green DB, Shaffer HB, Kats LB. 2016 Individual fluctuations in toxin levels affect breeding site fidelity in a chemically defended amphibian. Proc. R. Soc. B 283: 20160468. http://dx.doi.org/10.1098/rspb.2016.0468
Received: 2 March 2016 Accepted: 19 April 2016
Subject Areas: behaviour, ecology, evolution Keywords: site fidelity, newts, Taricha, tetrodotoxin, coevolution, Santa Monica Mountains
Author for correspondence: Gary M. Bucciarelli e-mail: [email protected]
Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2016.0468 or via http://rspb.royalsocietypublishing.org.
Individual fluctuations in toxin levels affect breeding site fidelity in a chemically defended amphibian Gary M. Bucciarelli1,2, David B. Green3, H. Bradley Shaffer1,2 and Lee B. Kats3 1
Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA 90095, USA UCLA La Kretz Center for California Conservation Science, Institute of the Environment and Sustainability, Los Angeles, CA 90095, USA 3 Natural Science Division, Pepperdine University, 24255 Pacific Coast Highway, Malibu, CA 90263, USA 2
Behaviours that influence habitat selection strongly determine species movement patterns. One component of animal behaviour that largely influences movement patterns and habitat choice is site fidelity. California newts (family Salamandridae) demonstrate remarkable site fidelity, typically homing to the same pool of a stream each breeding season. Individuals often occupy a specific pool throughout the breeding season, but some males shift among breeding pools, altering their set of potential mates, competitors, and predators. In this study, we measured dermal concentrations of the chemical defence compound tetrodotoxin (TTX) in recaptured male California newts (Taricha torosa) over five breeding seasons to evaluate whether relative TTX concentrations are associated with breeding site fidelity in the field. Our five years of field sampling indicates that TTX concentrations of individuals and group means fluctuate tremendously, implying that TTX is not a stable phenotypic trait. Despite such fluctuations, we found that an individual’s relative TTX concentration explains fidelity to a breeding pool and suggests that newts may be able to assess both their own concentrations of TTX and that of conspecifics to make decisions about remaining in or abandoning a breeding pool. These results provide us a novel dimension to chemical defence phenotypes in nature and their ecological consequences, potentially requiring a re-evaluation of the coevolutionary dynamics of predation pressure on toxin-laden organisms.
1. Introduction The constraints and mechanisms by which species select their breeding habitat is fundamental to their ecology, evolution, and conservation. The behaviours that influence habitat selection, for example, strongly determine species movement patterns, thereby affecting interactions with other organisms and the environment [1], fitness and local density [2], the spatial scale at which populations are regulated [3], metapopulation dynamics [4], and the evolutionary units within a population [5]. One key behaviour that largely influences movement patterns and habitat choice is site fidelity. Site fidelity, whereby individuals tend to return to a previously occupied site, is a life-history strategy that occurs in many species of three phyla; chordata, arthropoda, and mollusca [6]. The behaviour may increase survival by allowing animals to efficiently relocate and use previously successful sites for breeding, feeding, and overwintering [7], as well as reducing relative energetic costs [8], vulnerability to predators [9], and risk of mortality [10] associated with moving to new areas. Although site fidelity is a behaviour that links individual- and population-level processes with demonstrated ecological and evolutionary consequences [11], the mechanisms by which it occurs remain poorly understood in most systems. Some amphibians show remarkable site fidelity [12,13], and a considerable research effort spanning the last half-century has characterized its ecological consequences and underlying mechanisms. Some amphibian species are
& 2016 The Author(s) Published by the Royal Society. All rights reserved.
2. Methods (a) Temporal patterns of individual breeding site fidelity The breeding site history and fidelity of newts was reconstructed for two datasets of previously marked newts that were recaptured and sampled for TTX from 2011 to 2015 (electronic supplementary material, text). The first dataset included marked individuals sampled multiple times for TTX from 2011 to 2015 (n ¼ 32). The second dataset consisted of marked newts that were sampled for TTX only once during 2011–2015 (n ¼ 68). To determine the previous location of these newts prior to the first time they were sampled for TTX, we relied on a long-term mark–recapture survey dataset of 1 650 newts in Cold Creek that detailed where individuals were marked and recaptured from 1991 to 2015. Thus, every individual included in our analyses was previously caught, marked, and georeferenced at some point during 1991– 2015. Using these data, we determined if the first TTX sampling location occurred in the same pool where the male had previously been captured. For newts that were sampled more than once, we referred to the previous sampling location to determine if the subsequent sample occurred in the same pool. We then used a binary variable to indicate whether the newt was found in the same pool (0) or had moved (1) at the time of sampling. These binary data were used as the categorical response variable in our separate analyses of these two datasets. For example, a previously marked newt sampled in 2011 from pool A, then 2012 from pool B, and 2014 from pool A would first be searched for in our survey dataset to
2
Proc. R. Soc. B 283: 20160468
newts (Taricha torosa) causes conspecific larvae to flee and seek refuge [29,42]. TTX also affects the foraging behaviour of stream macroinvertebrate community members, and reduces movement and strike velocities of at least one macroinvertebrate predator [43]. In this study, we quantified TTX levels among individuals of a wild California newt population (T. torosa), and explored the relationship between TTX and site fidelity. We conducted longitudinal sampling of individual newts and used a multiyear mark–recapture study to track each individual’s TTX concentrations and movement patterns within a stream across multiple years. We place these data into the context of an ongoing, multi-decadal study of newt population biology in southern California streams during a severe drought and a time of precipitous amphibian declines. Overall, our results demonstrate that site fidelity at the stream level is exceptionally strong, with no marked individual recaptured in adjacent streams over 20 years of intensive study and hundreds of marked and recaptured newts. However, consistent with other studies, we have observed that some individuals will move up- or downstream among breeding pools or occupy different pools across years, and we analyse these temporal movement patterns as a consequence of newt morphology and an individual’s relative TTX concentrations. Given that TTX is both a defence compound and conspecific chemical cue, we designed our study to determine whether individual relative TTX concentrations influence site fidelity patterns in breeding male California newts. Our results indicate that (i) TTX levels of wild individuals fluctuate widely over time, (ii) mean TTX levels of our single breeding population span nearly the full range of T. torosa chemical defence levels previously measured across the entire species range, and (iii) adult relative TTX levels are linked to patterns of fidelity at breeding sites.
rspb.royalsocietypublishing.org
known to return year after year to the same pool or physical structures at a breeding site, with individuals often remaining in that microhabitat during the entire breeding period [14,15]. Other individuals may inexplicably abandon a breeding site, complicating explanations for the evolution and maintenance of site fidelity more generally. Olfaction, as an orientation mechanism, is critical in guiding site fidelity [12,16]. Olfactory cues influence many amphibian behaviours [17] and often play an important role in the recognition of individuals, kin, and mates [18,19]. Salamanders use chemical cues to recognize and select habitat [20,21], mark territories [22], and detect oviposition site quality [23]. Scent marking, which is typically associated with territorial behaviour [22], may also offer intruders the means to assess the competitive quality of other males [24]. Although olfaction can guide amphibian site fidelity, the proximate underpinnings of movement behaviour once individuals reach a breeding site remain poorly understood and difficult to study in the field. Determining the processes that facilitate individuals returning to and maintaining a breeding locality, and why they may abandon these sites, will help to inform questions surrounding the phenomenon of site fidelity. Newts of the genus Taricha are iteroparous, returning each spring to the same pond or reach of a stream to breed [12,25,26]. Newts are terrestrial most of the year and streambreeding Taricha navigate 250–500 m or more between terrestrial habitats and breeding sites [27] with pinpoint precision, often entering their stream within 15 m of previous years [12,26]. Throughout the breeding season, individuals generally remain within a few metres of their breeding pool, and often do not migrate after entering the stream [25]. Olfactory cues aid terrestrial migration to breeding sites [28], and once there, chemical cues remain critical components for sexual and feeding behaviour [29]. All four currently recognized species of Taricha possess a neurotoxin, tetrodotoxin (TTX), that is considered to have evolved primarily as a chemical defence. Population variation of Taricha TTX concentrations has been observed across broad geographical scales [30], a pattern that is presumed to be the result of a coevolutionary arms race between newts and predatory Thamnophis garter snakes [31]. The prevailing hypothesis, whereby Taricha TTX levels escalate in response to Thamnophis resistance to TTX, rests, in part, on the assumption that the TTX phenotype is genetically based and evolutionarily capable of responding to natural selection. However, the means by which Taricha produce and maintain TTX is unknown. Because TTX is an alkaloid, it is presumably not the direct product of a specific gene or gene family. The TTX concentrations of the sister taxon of Taricha, Notophthalmus newts from eastern North America, also vary across broad geographical ranges [32], but with no apparent relationship to a predator [33]. Furthermore, it appears that other tetrodotoxic taxa produce TTX via endosymbiotic bacteria rather than via an endogenous biosynthetic or genetically controlled pathway [34]. Given the extensive variation observed between individuals and populations [35], the role of TTX may well be more complex than that predicted purely by predator–prey interactions [36–38]. TTX appears to have selected for increased resistance in Thamnophis snakes [39], but it also appears to facilitate a diverse set of ecological processes. Various taxa rely on TTX as a feeding stimulant [40], sexual attractant [41], or antipredator chemical cue [29,42,43]. Specific to our study system, TTX from potentially cannibalistic adult California
We performed a series of analyses with classification models, first on our dataset composed of 32 individuals sampled for TTX multiple times, and then repeated the analysis with the data for the 68 individuals that were sampled for TTX only once. In general, the modelling approach evaluated relationships between T. torosa breeding pool fidelity and individual body condition, temporal variables, and TTX deviation (TTX), which is a measure of how much an individual’s TTX concentration on a sampling date deviated from that day’s mean TTX value of all newts sampled on that day (electronic supplementary material, S1 text). Thus, each individual sampled more than once had multiple TTX values, either positive or negative, for each date that they were sampled. Our decision to use an individual’s deviation from the mean on the day of sampling instead of absolute TTX concentrations was based on preliminary data that indicated individual and population concentrations fluctuated, which suggested there may be variation in the capacity of a chemical defence response at the individual level that could affect ecological or behavioural processes. Morphological predictors were obtained from individual measurements performed at each sampling event. The first was body condition (body), the ratio of mass to snout– vent length (SVL) and the second was measured tail height (tail), which we included because it is a secondary sexual characteristic and an indicator of fitness [44]. The temporal variables included in our models consisted of the year of sampling (year), month of sampling (month), time between sampling events (between) to account for potential lag effects, and a categorical variable (measure), which indicates the cumulative number of samples collected from an individual during the five years. The predictor measure was incorporated to detect a potential effect of repeated sampling on TTX levels of the 32 newts because they were sampled more than once. The first model implemented tree regression (TREE v. 1.0 – 29) [45] in R to identify predictors associated with breeding pool fidelity at Cold Creek. Tree regression uses a binary recursive partitioning method to construct a classification tree, hierarchically ranking predictor variables according to how much and at
3. Results (a) Sampling effort The 2011–2015 surveying and sampling effort was heavily male-biased, consisting of 465 encounters with males and 78 encounters with females. Of the total number of TTX samples collected during the sampling period, 73 of them were from 32 marked males repeatedly captured and sampled multiple times. An additional 68 samples came from marked males captured and sampled only once. The remaining number of TTX samples came from male newts that were not marked.
(b) Individual and group mean TTX concentrations During the 2011–2015 sampling period, a total of 231 dorsolateral tissue samples were collected from breeding males to subsequently quantitate TTX concentrations within Cold Creek. Mean TTX concentrations through time are shown in figure 1. Individual male TTX concentrations ranged from
3
Proc. R. Soc. B 283: 20160468
(b) Statistical methods
what level they explain variance in the response variable. This modelling approach is attractive because it is easily interpretable and indicates directionality and strength of a response. For our focal dataset of 32 newts, the tree regression model included the categorical temporal predictors year, month, and between; the morphological continuous variables body and tail; the predictor measure; the values for relative TTX concentrations, TTX; and the binary fidelity data, which were coded as the response variables. When analysing the dataset of 68 newts that were sampled only once, the predictors between and measure were deleted from our models because these males were only sampled once for TTX. To better understand which variables predict male fidelity to a breeding pool, we constructed a second model that applied tree regressions using the RANDOMFOREST framework (v. 4.6-6) [46] in R. This algorithm iteratively and randomly removes variables to evaluate relationships between predictor and response variables. For both datasets, we constructed the models to perform 10 000 iterations and used the same predictors as in our tree regression models. We used a random forest model rather than more traditional linear regression methods because it is a non-parametric procedure that does not assume an underlying distribution of the data. It has also been shown to outperform traditional regression tests [47], to adequately assess clustered datasets with repeated measures [48], and to describe the role of continuous variables across a landscape [49]. Another benefit of random forest models is their ability to handle complex interactions between variables [50]. To assess the robustness of our models, we used the predict function in R and data from our focal set of 32 newts to generate predicted values of the number of instances individuals would move or show fidelity. These predicted and the observed values from the dataset of 32 newts were statistically compared using a chi-square test in a 2 2 contingency table. In addition, p-values for each predictor were estimated in R with the package rfPermute (v. 1.6.2) by producing a null distribution of 1 000 permutation replicates of the random forest model that was used for the analysis of the 32 newts sampled multiple times. Finally, we computed matrices of Spearman’s rank correlations coefficients using the Hmisc (v. 3.9-3) framework [51] and rcorr function in R to test for correlations between TTX concentrations and measured individual morphological traits, which included mass, SVL, and tail height. We performed this analysis using our dataset of 32 individuals to determine whether greater individual TTX concentrations were associated with males having greater mass, size, or tail height.
rspb.royalsocietypublishing.org
determine where it was last found before 2011. If that male were last found in pool A in 2010, then the response variable would be coded as a ‘0’ in 2011, a ‘1’ in 2012, and a ‘1’ in 2014. Additionally, morphological and temporal data were included for each individual for each sampling event. Details of the morphological and temporal data are outlined under our Statistical methods section. For the 68 individuals that were sampled only once, we realize that this set of individuals does not provide the same picture of temporal variation of chemical defence data and patterns of site fidelity as do individuals sampled two or more times. However, they nonetheless allow us to use a much larger sample size to evaluate the relationship between movement, morphological predictors, and TTX concentrations. We also used the long-term mark – recapture survey data to understand breeding site fidelity at the population level. We compared recapture locations of these 1 650 newts and estimated the percentage of individuals recaptured in the same pool compared with those that were recaptured in different pools. To ensure that these two datasets are independent, we included only individuals recaptured, but never sampled for TTX between 1991 and 2015. Ultimately, these data provide us a much larger temporal context for our five years of TTX and movement data, thereby allowing us to determine whether the observed movement patterns of sampled individuals among pools from 2011 to 2015 was extraordinary relative to movement over the previous two decades.
4
2.8
rspb.royalsocietypublishing.org
nanomoles TTX mg–1 skin
3.2
2.4 2.0 1.6 1.2 0.8 0.4 M AM J J A 2011
M A M J 2012
M A M J 2013
M A M J 2014
M A M J 2015
M A M J J A 2011
M A M J 2012
M A M J 2013
M A M J 2014
M A M J 2015
2.00 1.75
TTX (mg)
1.50 1.25 1.00 0.75 0.50 0.25 0
Figure 1. Fluctuations in mean TTX concentrations derived from adult male T. torosa in Cold Creek. Diamonds show mean male TTX concentrations (+s.e.) for each sampling date during 2011– 2015. The measured TTX concentrations are presented in units of nanomoles TTX mg21 skin [52] and estimated whole individual TTX concentrations in mg TTX [53]. Coloured segments in the bottom graph illustrate the scale of ‘whole newt toxicity’ from Hanifin et al. [53]. Regardless of units, mean TTX levels within this breeding population fluctuate nearly the full scale previously used to describe patterns of phenotypic variation across the entire species range. 0.00 to 5.72 nmol TTX mg21 skin (electronic supplementary material, table S1). Daily mean TTX values varied within a season and between years, fluctuating from a high of 2.22 nmol TTX mg21 on 15 May 2012, to a low of 0.03 nmol TTX mg21 on 24 April 2013 (figure 1 and electronic supplementary material, table S2). Overall, the most extreme individual TTX deviation values were from newts sampled in March 2011 (þ3.60) and May 2012 (22.04).
(c) Temporal patterns of individual breeding site fidelity All 32 individuals that were sampled multiple times for TTX were found in the 10 largest pools within Cold Creek, except one that was recaptured once in an intervening side pool (figure 2). Fidelity to a breeding pool was common year after year for the majority of individuals. Of the 32 individuals that were recaptured, approximately 68% were always recaptured in the same pool. Similarly, of the 68 individuals sampled for TTX only once, 42 individuals (62%) were recaptured in the same pool where they were originally captured. Finally, using the survey data collected from 1991–2015, we found that approximately 72% of newts were recaptured in the same pool across years.
(d) Movement among pools Tree regression indicated that fidelity to a breeding pool is best explained by TTX deviation values and morphology (figure 3a,c). The most important correlate for both sets of data was an individual’s relative TTX level (TTX), such that
recaptures repeatedly occupying the same pool were associated with positive TTX deviation values. In our focal set of 32 individuals, males with deviation values greater than þ0.38 nmol TTX mg21 were recaptured in the same breeding pool. For individuals with TTX deviation values less than þ0.38 nmol TTX mg21, fidelity was explained by a combination of both tail height and deviation values. Any male with a TTX deviation value less than þ0.38 nmol TTX mg21 and a tail that was less than 11.9 mm tended to move, whereas when tail height exceeded 11.9 mm, the fidelity of those males was explained further by TTX values. These males with taller tails tended to move if their TTX value was less than 20.28 nmol TTX mg21 and show fidelity if TTX values exceeded 20.28 nmol TTX mg21 (figure 3a). Similarly, a deviation value of þ0.12 nmol TTX mg21 best explained fidelity patterns in the 68 individuals sampled only once (figure 3c). Body condition was the next most important predictor of movement for this dataset of newts sampled only once. Overall, newts with below average TTX concentrations and shorter tails or low body condition ratios were often recaptured in different breeding pools over time. The random forest models showed that individual TTX values and morphological predictors consistently explained fidelity to a breeding pool more than time related predictors (figure 3b,d). As with the tree regression models, TTX was the most important predictor, followed by tail height (tail) and body condition (body). Predicted versus observed values did not significantly differ from one another (x 2, d.f. ¼ 1, p ¼ 0.84). All estimated
Proc. R. Soc. B 283: 20160468
0
>+3
TTX deviation value >–3 0
5
breeding pools
individuals showing fidelity
individuals not showing fidelity n i cc dd ee
h
u w g
k l e r s
a ff t b o v
y c z aa
x
2011
2012
2013
2014
2015
March April May June July
Figure 2. Individuals with above average TTX levels tend to show fidelity to a breeding pool. Coloured circles show individual TTX deviation values, a calculation of how much an individual’s TTX concentration on a sampling date deviated from the mean TTX concentration of all newts sampled on that same day. Individuals recaptured in the same pool (left side) over our five-year study tend to have greater positive TTX values compared with individuals recaptured in different pools (right side). The vertical segmented bar represents the length of the study site in the Cold Creek Preserve, with each alternating segment scaled to a 20 m section of the stream (note the break in scale). Each horizontal dashed line represents an individual newt (identified by letter, see electronic supplementary material, table S1) and the five-sampling period. Arrows point to the breeding pool(s) (grey circles) where an individual was recaptured. Using the values from tree regression results, only measurements and individuals with deviation values exceeding þ0.38 and less than 20.28 are plotted. p-values generated in rfPermute for every predictor exceeded the 0.05 a level except the predictor TTX ( p ¼ 0.001), which was our measure of relative TTX concentrations (electronic supplementary material, figure S1). Test statistics from Spearman’s rank analyses indicated that no morphological trait was significantly correlated with TTX concentrations (mass: p ¼ 0.29; SVL: p ¼ 0.49; tail height: p ¼ 0.99).
4. Discussion Although breeding site fidelity is a well-documented phenomenon in a number of amphibian species, a more proximate understanding of why some individuals repeatedly use the same site and others do not remains obscure. In a few cases where it has been studied, individual attributes have been identified as a potential causal explanation. For example, Kopecky´ et al. [54] observed that male Alpine newts (Mesotrition alpestris) with high body condition tended to move among adjacent, human-created breeding sites, perhaps, because such movement enhances the most fit male’s ability to gain increased access to mates. However, most studies simply
document the fraction of individuals that move among breeding sites [55]. If we are to understand why some individuals have strong site fidelity and others do not, and use that information to better understand metapopulation dynamics and conserve declining species as land fragmentation and loss of aquatic habitat occur, a more mechanistic grasp of site fidelity is necessary. In our study, an individual’s relative TTX concentrations best explained site fidelity patterns, such that males with low TTX levels relative to other males concurrently occupying the stream tended to be recaptured in different pools. Roughly, two-thirds of newts were repeatedly recaptured in the same pool during our 2011–2015 study, while the remaining third moved among pools separated by more than 20 m. From our analyses, these movement patterns are strongly correlated with relative TTX concentrations. Tree regression and random forest models on both sets of recapture data indicate that temporal patterns do not explain these differences. Rather, relative TTX concentrations and morphological features best explain fidelity to a breeding pool. The similarity in model results for the two different datasets, which cover different temporal periods, provides strong evidence of the importance of TTX in site
Proc. R. Soc. B 283: 20160468
d
rspb.royalsocietypublishing.org
+0.38 –0.28
(a)
(b)
TTX
6
rspb.royalsocietypublishing.org
TTX tail body month year between >0.38 nmol TTX mg–1 fidelity
tail
measure
(c)
Proc. R. Soc. B 283: 20160468
0 2 4 6 8 10 12 mean decrease in accuracy (Gini impurity score)
0.12 nmol TTX fidelity
body
month
mg–1
0 2 4 6 8 mean decrease in accuracy (Gini impurity score)
0.16 mas:SVL fidelity moves
Figure 3. Tree and random forest model results indicate relative TTX levels are most important in explaining fidelity. (a,c) Tree regressions depict the relationship between breeding pool fidelity and individual’s relative TTX concentrations and morphology for the focal set of 32 newts sampled multiple times (a) and the 68 individuals sampled once (c). Overall, the results indicate that males with greater relative TTX concentrations tend to remain in the same pools. Variables in the random forest models are shown for the 32 newts sampled multiple times (b) and only once (d ), ordered top-to-bottom from most to least important in explaining fidelity. Both random forest models indicate that relative TTX is the most important predictor in explaining breeding site fidelity, followed by morphological variables. Hanifin et al. [53]
2011 annual mean TTX (mg) metres
2 200
N
0
500
2012 annual mean TTX (mg) metres
2 200
0
0
metres
2 200 500
0
2015 annual mean TTX (mg) 0
2 200
2 200
2 200
1 600
1 600
1 600
1 600
1 600
M
A
M
J
J
monthly mean TTX (mg) monthly mean TTX (mg) monthly mean TTX (mg)
M
A
M
J
RD
J
NT U
J
500
ST
M
RD
A
NT
M
U ST
J
RD
J
NT U
M
RD
RD
A
ST
NT U
ST
NT U
ST
M
metres
2 200 500
2 200
0.3 – 0.5
newt toxin index (mg)
metres
2 200
2014 annual mean TTX (mg)
2 200
>2.0
Research Cite this article: Bucciarelli GM, Green DB, Shaffer HB, Kats LB. 2016 Individual fluctuations in toxin levels affect breeding site fidelity in a chemically defended amphibian. Proc. R. Soc. B 283: 20160468. http://dx.doi.org/10.1098/rspb.2016.0468
Received: 2 March 2016 Accepted: 19 April 2016
Subject Areas: behaviour, ecology, evolution Keywords: site fidelity, newts, Taricha, tetrodotoxin, coevolution, Santa Monica Mountains
Author for correspondence: Gary M. Bucciarelli e-mail: [email protected]
Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2016.0468 or via http://rspb.royalsocietypublishing.org.
Individual fluctuations in toxin levels affect breeding site fidelity in a chemically defended amphibian Gary M. Bucciarelli1,2, David B. Green3, H. Bradley Shaffer1,2 and Lee B. Kats3 1
Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA 90095, USA UCLA La Kretz Center for California Conservation Science, Institute of the Environment and Sustainability, Los Angeles, CA 90095, USA 3 Natural Science Division, Pepperdine University, 24255 Pacific Coast Highway, Malibu, CA 90263, USA 2
Behaviours that influence habitat selection strongly determine species movement patterns. One component of animal behaviour that largely influences movement patterns and habitat choice is site fidelity. California newts (family Salamandridae) demonstrate remarkable site fidelity, typically homing to the same pool of a stream each breeding season. Individuals often occupy a specific pool throughout the breeding season, but some males shift among breeding pools, altering their set of potential mates, competitors, and predators. In this study, we measured dermal concentrations of the chemical defence compound tetrodotoxin (TTX) in recaptured male California newts (Taricha torosa) over five breeding seasons to evaluate whether relative TTX concentrations are associated with breeding site fidelity in the field. Our five years of field sampling indicates that TTX concentrations of individuals and group means fluctuate tremendously, implying that TTX is not a stable phenotypic trait. Despite such fluctuations, we found that an individual’s relative TTX concentration explains fidelity to a breeding pool and suggests that newts may be able to assess both their own concentrations of TTX and that of conspecifics to make decisions about remaining in or abandoning a breeding pool. These results provide us a novel dimension to chemical defence phenotypes in nature and their ecological consequences, potentially requiring a re-evaluation of the coevolutionary dynamics of predation pressure on toxin-laden organisms.
1. Introduction The constraints and mechanisms by which species select their breeding habitat is fundamental to their ecology, evolution, and conservation. The behaviours that influence habitat selection, for example, strongly determine species movement patterns, thereby affecting interactions with other organisms and the environment [1], fitness and local density [2], the spatial scale at which populations are regulated [3], metapopulation dynamics [4], and the evolutionary units within a population [5]. One key behaviour that largely influences movement patterns and habitat choice is site fidelity. Site fidelity, whereby individuals tend to return to a previously occupied site, is a life-history strategy that occurs in many species of three phyla; chordata, arthropoda, and mollusca [6]. The behaviour may increase survival by allowing animals to efficiently relocate and use previously successful sites for breeding, feeding, and overwintering [7], as well as reducing relative energetic costs [8], vulnerability to predators [9], and risk of mortality [10] associated with moving to new areas. Although site fidelity is a behaviour that links individual- and population-level processes with demonstrated ecological and evolutionary consequences [11], the mechanisms by which it occurs remain poorly understood in most systems. Some amphibians show remarkable site fidelity [12,13], and a considerable research effort spanning the last half-century has characterized its ecological consequences and underlying mechanisms. Some amphibian species are
& 2016 The Author(s) Published by the Royal Society. All rights reserved.
2. Methods (a) Temporal patterns of individual breeding site fidelity The breeding site history and fidelity of newts was reconstructed for two datasets of previously marked newts that were recaptured and sampled for TTX from 2011 to 2015 (electronic supplementary material, text). The first dataset included marked individuals sampled multiple times for TTX from 2011 to 2015 (n ¼ 32). The second dataset consisted of marked newts that were sampled for TTX only once during 2011–2015 (n ¼ 68). To determine the previous location of these newts prior to the first time they were sampled for TTX, we relied on a long-term mark–recapture survey dataset of 1 650 newts in Cold Creek that detailed where individuals were marked and recaptured from 1991 to 2015. Thus, every individual included in our analyses was previously caught, marked, and georeferenced at some point during 1991– 2015. Using these data, we determined if the first TTX sampling location occurred in the same pool where the male had previously been captured. For newts that were sampled more than once, we referred to the previous sampling location to determine if the subsequent sample occurred in the same pool. We then used a binary variable to indicate whether the newt was found in the same pool (0) or had moved (1) at the time of sampling. These binary data were used as the categorical response variable in our separate analyses of these two datasets. For example, a previously marked newt sampled in 2011 from pool A, then 2012 from pool B, and 2014 from pool A would first be searched for in our survey dataset to
2
Proc. R. Soc. B 283: 20160468
newts (Taricha torosa) causes conspecific larvae to flee and seek refuge [29,42]. TTX also affects the foraging behaviour of stream macroinvertebrate community members, and reduces movement and strike velocities of at least one macroinvertebrate predator [43]. In this study, we quantified TTX levels among individuals of a wild California newt population (T. torosa), and explored the relationship between TTX and site fidelity. We conducted longitudinal sampling of individual newts and used a multiyear mark–recapture study to track each individual’s TTX concentrations and movement patterns within a stream across multiple years. We place these data into the context of an ongoing, multi-decadal study of newt population biology in southern California streams during a severe drought and a time of precipitous amphibian declines. Overall, our results demonstrate that site fidelity at the stream level is exceptionally strong, with no marked individual recaptured in adjacent streams over 20 years of intensive study and hundreds of marked and recaptured newts. However, consistent with other studies, we have observed that some individuals will move up- or downstream among breeding pools or occupy different pools across years, and we analyse these temporal movement patterns as a consequence of newt morphology and an individual’s relative TTX concentrations. Given that TTX is both a defence compound and conspecific chemical cue, we designed our study to determine whether individual relative TTX concentrations influence site fidelity patterns in breeding male California newts. Our results indicate that (i) TTX levels of wild individuals fluctuate widely over time, (ii) mean TTX levels of our single breeding population span nearly the full range of T. torosa chemical defence levels previously measured across the entire species range, and (iii) adult relative TTX levels are linked to patterns of fidelity at breeding sites.
rspb.royalsocietypublishing.org
known to return year after year to the same pool or physical structures at a breeding site, with individuals often remaining in that microhabitat during the entire breeding period [14,15]. Other individuals may inexplicably abandon a breeding site, complicating explanations for the evolution and maintenance of site fidelity more generally. Olfaction, as an orientation mechanism, is critical in guiding site fidelity [12,16]. Olfactory cues influence many amphibian behaviours [17] and often play an important role in the recognition of individuals, kin, and mates [18,19]. Salamanders use chemical cues to recognize and select habitat [20,21], mark territories [22], and detect oviposition site quality [23]. Scent marking, which is typically associated with territorial behaviour [22], may also offer intruders the means to assess the competitive quality of other males [24]. Although olfaction can guide amphibian site fidelity, the proximate underpinnings of movement behaviour once individuals reach a breeding site remain poorly understood and difficult to study in the field. Determining the processes that facilitate individuals returning to and maintaining a breeding locality, and why they may abandon these sites, will help to inform questions surrounding the phenomenon of site fidelity. Newts of the genus Taricha are iteroparous, returning each spring to the same pond or reach of a stream to breed [12,25,26]. Newts are terrestrial most of the year and streambreeding Taricha navigate 250–500 m or more between terrestrial habitats and breeding sites [27] with pinpoint precision, often entering their stream within 15 m of previous years [12,26]. Throughout the breeding season, individuals generally remain within a few metres of their breeding pool, and often do not migrate after entering the stream [25]. Olfactory cues aid terrestrial migration to breeding sites [28], and once there, chemical cues remain critical components for sexual and feeding behaviour [29]. All four currently recognized species of Taricha possess a neurotoxin, tetrodotoxin (TTX), that is considered to have evolved primarily as a chemical defence. Population variation of Taricha TTX concentrations has been observed across broad geographical scales [30], a pattern that is presumed to be the result of a coevolutionary arms race between newts and predatory Thamnophis garter snakes [31]. The prevailing hypothesis, whereby Taricha TTX levels escalate in response to Thamnophis resistance to TTX, rests, in part, on the assumption that the TTX phenotype is genetically based and evolutionarily capable of responding to natural selection. However, the means by which Taricha produce and maintain TTX is unknown. Because TTX is an alkaloid, it is presumably not the direct product of a specific gene or gene family. The TTX concentrations of the sister taxon of Taricha, Notophthalmus newts from eastern North America, also vary across broad geographical ranges [32], but with no apparent relationship to a predator [33]. Furthermore, it appears that other tetrodotoxic taxa produce TTX via endosymbiotic bacteria rather than via an endogenous biosynthetic or genetically controlled pathway [34]. Given the extensive variation observed between individuals and populations [35], the role of TTX may well be more complex than that predicted purely by predator–prey interactions [36–38]. TTX appears to have selected for increased resistance in Thamnophis snakes [39], but it also appears to facilitate a diverse set of ecological processes. Various taxa rely on TTX as a feeding stimulant [40], sexual attractant [41], or antipredator chemical cue [29,42,43]. Specific to our study system, TTX from potentially cannibalistic adult California
We performed a series of analyses with classification models, first on our dataset composed of 32 individuals sampled for TTX multiple times, and then repeated the analysis with the data for the 68 individuals that were sampled for TTX only once. In general, the modelling approach evaluated relationships between T. torosa breeding pool fidelity and individual body condition, temporal variables, and TTX deviation (TTX), which is a measure of how much an individual’s TTX concentration on a sampling date deviated from that day’s mean TTX value of all newts sampled on that day (electronic supplementary material, S1 text). Thus, each individual sampled more than once had multiple TTX values, either positive or negative, for each date that they were sampled. Our decision to use an individual’s deviation from the mean on the day of sampling instead of absolute TTX concentrations was based on preliminary data that indicated individual and population concentrations fluctuated, which suggested there may be variation in the capacity of a chemical defence response at the individual level that could affect ecological or behavioural processes. Morphological predictors were obtained from individual measurements performed at each sampling event. The first was body condition (body), the ratio of mass to snout– vent length (SVL) and the second was measured tail height (tail), which we included because it is a secondary sexual characteristic and an indicator of fitness [44]. The temporal variables included in our models consisted of the year of sampling (year), month of sampling (month), time between sampling events (between) to account for potential lag effects, and a categorical variable (measure), which indicates the cumulative number of samples collected from an individual during the five years. The predictor measure was incorporated to detect a potential effect of repeated sampling on TTX levels of the 32 newts because they were sampled more than once. The first model implemented tree regression (TREE v. 1.0 – 29) [45] in R to identify predictors associated with breeding pool fidelity at Cold Creek. Tree regression uses a binary recursive partitioning method to construct a classification tree, hierarchically ranking predictor variables according to how much and at
3. Results (a) Sampling effort The 2011–2015 surveying and sampling effort was heavily male-biased, consisting of 465 encounters with males and 78 encounters with females. Of the total number of TTX samples collected during the sampling period, 73 of them were from 32 marked males repeatedly captured and sampled multiple times. An additional 68 samples came from marked males captured and sampled only once. The remaining number of TTX samples came from male newts that were not marked.
(b) Individual and group mean TTX concentrations During the 2011–2015 sampling period, a total of 231 dorsolateral tissue samples were collected from breeding males to subsequently quantitate TTX concentrations within Cold Creek. Mean TTX concentrations through time are shown in figure 1. Individual male TTX concentrations ranged from
3
Proc. R. Soc. B 283: 20160468
(b) Statistical methods
what level they explain variance in the response variable. This modelling approach is attractive because it is easily interpretable and indicates directionality and strength of a response. For our focal dataset of 32 newts, the tree regression model included the categorical temporal predictors year, month, and between; the morphological continuous variables body and tail; the predictor measure; the values for relative TTX concentrations, TTX; and the binary fidelity data, which were coded as the response variables. When analysing the dataset of 68 newts that were sampled only once, the predictors between and measure were deleted from our models because these males were only sampled once for TTX. To better understand which variables predict male fidelity to a breeding pool, we constructed a second model that applied tree regressions using the RANDOMFOREST framework (v. 4.6-6) [46] in R. This algorithm iteratively and randomly removes variables to evaluate relationships between predictor and response variables. For both datasets, we constructed the models to perform 10 000 iterations and used the same predictors as in our tree regression models. We used a random forest model rather than more traditional linear regression methods because it is a non-parametric procedure that does not assume an underlying distribution of the data. It has also been shown to outperform traditional regression tests [47], to adequately assess clustered datasets with repeated measures [48], and to describe the role of continuous variables across a landscape [49]. Another benefit of random forest models is their ability to handle complex interactions between variables [50]. To assess the robustness of our models, we used the predict function in R and data from our focal set of 32 newts to generate predicted values of the number of instances individuals would move or show fidelity. These predicted and the observed values from the dataset of 32 newts were statistically compared using a chi-square test in a 2 2 contingency table. In addition, p-values for each predictor were estimated in R with the package rfPermute (v. 1.6.2) by producing a null distribution of 1 000 permutation replicates of the random forest model that was used for the analysis of the 32 newts sampled multiple times. Finally, we computed matrices of Spearman’s rank correlations coefficients using the Hmisc (v. 3.9-3) framework [51] and rcorr function in R to test for correlations between TTX concentrations and measured individual morphological traits, which included mass, SVL, and tail height. We performed this analysis using our dataset of 32 individuals to determine whether greater individual TTX concentrations were associated with males having greater mass, size, or tail height.
rspb.royalsocietypublishing.org
determine where it was last found before 2011. If that male were last found in pool A in 2010, then the response variable would be coded as a ‘0’ in 2011, a ‘1’ in 2012, and a ‘1’ in 2014. Additionally, morphological and temporal data were included for each individual for each sampling event. Details of the morphological and temporal data are outlined under our Statistical methods section. For the 68 individuals that were sampled only once, we realize that this set of individuals does not provide the same picture of temporal variation of chemical defence data and patterns of site fidelity as do individuals sampled two or more times. However, they nonetheless allow us to use a much larger sample size to evaluate the relationship between movement, morphological predictors, and TTX concentrations. We also used the long-term mark – recapture survey data to understand breeding site fidelity at the population level. We compared recapture locations of these 1 650 newts and estimated the percentage of individuals recaptured in the same pool compared with those that were recaptured in different pools. To ensure that these two datasets are independent, we included only individuals recaptured, but never sampled for TTX between 1991 and 2015. Ultimately, these data provide us a much larger temporal context for our five years of TTX and movement data, thereby allowing us to determine whether the observed movement patterns of sampled individuals among pools from 2011 to 2015 was extraordinary relative to movement over the previous two decades.
4
2.8
rspb.royalsocietypublishing.org
nanomoles TTX mg–1 skin
3.2
2.4 2.0 1.6 1.2 0.8 0.4 M AM J J A 2011
M A M J 2012
M A M J 2013
M A M J 2014
M A M J 2015
M A M J J A 2011
M A M J 2012
M A M J 2013
M A M J 2014
M A M J 2015
2.00 1.75
TTX (mg)
1.50 1.25 1.00 0.75 0.50 0.25 0
Figure 1. Fluctuations in mean TTX concentrations derived from adult male T. torosa in Cold Creek. Diamonds show mean male TTX concentrations (+s.e.) for each sampling date during 2011– 2015. The measured TTX concentrations are presented in units of nanomoles TTX mg21 skin [52] and estimated whole individual TTX concentrations in mg TTX [53]. Coloured segments in the bottom graph illustrate the scale of ‘whole newt toxicity’ from Hanifin et al. [53]. Regardless of units, mean TTX levels within this breeding population fluctuate nearly the full scale previously used to describe patterns of phenotypic variation across the entire species range. 0.00 to 5.72 nmol TTX mg21 skin (electronic supplementary material, table S1). Daily mean TTX values varied within a season and between years, fluctuating from a high of 2.22 nmol TTX mg21 on 15 May 2012, to a low of 0.03 nmol TTX mg21 on 24 April 2013 (figure 1 and electronic supplementary material, table S2). Overall, the most extreme individual TTX deviation values were from newts sampled in March 2011 (þ3.60) and May 2012 (22.04).
(c) Temporal patterns of individual breeding site fidelity All 32 individuals that were sampled multiple times for TTX were found in the 10 largest pools within Cold Creek, except one that was recaptured once in an intervening side pool (figure 2). Fidelity to a breeding pool was common year after year for the majority of individuals. Of the 32 individuals that were recaptured, approximately 68% were always recaptured in the same pool. Similarly, of the 68 individuals sampled for TTX only once, 42 individuals (62%) were recaptured in the same pool where they were originally captured. Finally, using the survey data collected from 1991–2015, we found that approximately 72% of newts were recaptured in the same pool across years.
(d) Movement among pools Tree regression indicated that fidelity to a breeding pool is best explained by TTX deviation values and morphology (figure 3a,c). The most important correlate for both sets of data was an individual’s relative TTX level (TTX), such that
recaptures repeatedly occupying the same pool were associated with positive TTX deviation values. In our focal set of 32 individuals, males with deviation values greater than þ0.38 nmol TTX mg21 were recaptured in the same breeding pool. For individuals with TTX deviation values less than þ0.38 nmol TTX mg21, fidelity was explained by a combination of both tail height and deviation values. Any male with a TTX deviation value less than þ0.38 nmol TTX mg21 and a tail that was less than 11.9 mm tended to move, whereas when tail height exceeded 11.9 mm, the fidelity of those males was explained further by TTX values. These males with taller tails tended to move if their TTX value was less than 20.28 nmol TTX mg21 and show fidelity if TTX values exceeded 20.28 nmol TTX mg21 (figure 3a). Similarly, a deviation value of þ0.12 nmol TTX mg21 best explained fidelity patterns in the 68 individuals sampled only once (figure 3c). Body condition was the next most important predictor of movement for this dataset of newts sampled only once. Overall, newts with below average TTX concentrations and shorter tails or low body condition ratios were often recaptured in different breeding pools over time. The random forest models showed that individual TTX values and morphological predictors consistently explained fidelity to a breeding pool more than time related predictors (figure 3b,d). As with the tree regression models, TTX was the most important predictor, followed by tail height (tail) and body condition (body). Predicted versus observed values did not significantly differ from one another (x 2, d.f. ¼ 1, p ¼ 0.84). All estimated
Proc. R. Soc. B 283: 20160468
0
>+3
TTX deviation value >–3 0
5
breeding pools
individuals showing fidelity
individuals not showing fidelity n i cc dd ee
h
u w g
k l e r s
a ff t b o v
y c z aa
x
2011
2012
2013
2014
2015
March April May June July
Figure 2. Individuals with above average TTX levels tend to show fidelity to a breeding pool. Coloured circles show individual TTX deviation values, a calculation of how much an individual’s TTX concentration on a sampling date deviated from the mean TTX concentration of all newts sampled on that same day. Individuals recaptured in the same pool (left side) over our five-year study tend to have greater positive TTX values compared with individuals recaptured in different pools (right side). The vertical segmented bar represents the length of the study site in the Cold Creek Preserve, with each alternating segment scaled to a 20 m section of the stream (note the break in scale). Each horizontal dashed line represents an individual newt (identified by letter, see electronic supplementary material, table S1) and the five-sampling period. Arrows point to the breeding pool(s) (grey circles) where an individual was recaptured. Using the values from tree regression results, only measurements and individuals with deviation values exceeding þ0.38 and less than 20.28 are plotted. p-values generated in rfPermute for every predictor exceeded the 0.05 a level except the predictor TTX ( p ¼ 0.001), which was our measure of relative TTX concentrations (electronic supplementary material, figure S1). Test statistics from Spearman’s rank analyses indicated that no morphological trait was significantly correlated with TTX concentrations (mass: p ¼ 0.29; SVL: p ¼ 0.49; tail height: p ¼ 0.99).
4. Discussion Although breeding site fidelity is a well-documented phenomenon in a number of amphibian species, a more proximate understanding of why some individuals repeatedly use the same site and others do not remains obscure. In a few cases where it has been studied, individual attributes have been identified as a potential causal explanation. For example, Kopecky´ et al. [54] observed that male Alpine newts (Mesotrition alpestris) with high body condition tended to move among adjacent, human-created breeding sites, perhaps, because such movement enhances the most fit male’s ability to gain increased access to mates. However, most studies simply
document the fraction of individuals that move among breeding sites [55]. If we are to understand why some individuals have strong site fidelity and others do not, and use that information to better understand metapopulation dynamics and conserve declining species as land fragmentation and loss of aquatic habitat occur, a more mechanistic grasp of site fidelity is necessary. In our study, an individual’s relative TTX concentrations best explained site fidelity patterns, such that males with low TTX levels relative to other males concurrently occupying the stream tended to be recaptured in different pools. Roughly, two-thirds of newts were repeatedly recaptured in the same pool during our 2011–2015 study, while the remaining third moved among pools separated by more than 20 m. From our analyses, these movement patterns are strongly correlated with relative TTX concentrations. Tree regression and random forest models on both sets of recapture data indicate that temporal patterns do not explain these differences. Rather, relative TTX concentrations and morphological features best explain fidelity to a breeding pool. The similarity in model results for the two different datasets, which cover different temporal periods, provides strong evidence of the importance of TTX in site
Proc. R. Soc. B 283: 20160468
d
rspb.royalsocietypublishing.org
+0.38 –0.28
(a)
(b)
TTX
6
rspb.royalsocietypublishing.org
TTX tail body month year between >0.38 nmol TTX mg–1 fidelity
tail
measure
(c)
Proc. R. Soc. B 283: 20160468
0 2 4 6 8 10 12 mean decrease in accuracy (Gini impurity score)
0.12 nmol TTX fidelity
body
month
mg–1
0 2 4 6 8 mean decrease in accuracy (Gini impurity score)
0.16 mas:SVL fidelity moves
Figure 3. Tree and random forest model results indicate relative TTX levels are most important in explaining fidelity. (a,c) Tree regressions depict the relationship between breeding pool fidelity and individual’s relative TTX concentrations and morphology for the focal set of 32 newts sampled multiple times (a) and the 68 individuals sampled once (c). Overall, the results indicate that males with greater relative TTX concentrations tend to remain in the same pools. Variables in the random forest models are shown for the 32 newts sampled multiple times (b) and only once (d ), ordered top-to-bottom from most to least important in explaining fidelity. Both random forest models indicate that relative TTX is the most important predictor in explaining breeding site fidelity, followed by morphological variables. Hanifin et al. [53]
2011 annual mean TTX (mg) metres
2 200
N
0
500
2012 annual mean TTX (mg) metres
2 200
0
0
metres
2 200 500
0
2015 annual mean TTX (mg) 0
2 200
2 200
2 200
1 600
1 600
1 600
1 600
1 600
M
A
M
J
J
monthly mean TTX (mg) monthly mean TTX (mg) monthly mean TTX (mg)
M
A
M
J
RD
J
NT U
J
500
ST
M
RD
A
NT
M
U ST
J
RD
J
NT U
M
RD
RD
A
ST
NT U
ST
NT U
ST
M
metres
2 200 500
2 200
0.3 – 0.5
newt toxin index (mg)
metres
2 200
2014 annual mean TTX (mg)
2 200
>2.0
